Altos Research: Leading Indicators of Home Prices (Sept 2010)

8/8/2019 Altos Research: Leading Indicators of Home Prices (Sept 2010)

1/44

An Altos Research White Paper

Active Housing Market Conditions

as

Home Price Leading Indicators

Forecasting Market Inflections

September, 2010


2/44

Active Housing Market Conditions as Home Price Leading Indicators

Copyright 2010 Altos Research LLCAll Rights Reserved 1

Executive Summary

Watching active housing market statistics for clues about the future prices of

homes.

Traditional housing market analysis orients on home price valuation based on closedcomps of moderately-recent, similar transactions. In 2007, this backward-facing view of thehousing market failed spectacularly.

It became obvious that to have accurate, real-time information about housing marketdynamics in different geographies and price ranges, and to be able to project with the highestpossible accuracy where those prices might be headed in the future was the new imperative.

Indices of home prices are aggregated, usually on a monthly or quarterly basis, from data setsof property and transaction information for sold residences as recorded at the county level.One problem for investors and traders is the necessary time lag involved in this process.Once a sale price is agreed to, it typically takes several weeks for the escrow to close and thesale to be reported to the county. For the county recorders office to enter the sale recordsinto its databaseand to complete the set of sales taking place across the county over a fullmonth introduces another time delay, sometimes measured in days or weeks andsometimes longer. After that, the data set must be analyzed to generate an index. Thus, thereis an unavoidable time lag for any index based on recorded sale prices.

As an example, the Standard & Poors Case-Shiller Home Price Indices are released on thelast Tuesday of every month with a two-month time lag. Their March 2010 indices werereleased on May 25, 2010. This time lag simply reflects the delays inherent in recording saleprices, assembling a complete monthly data set for a metropolitan area, and analyzing theresulting data to produce an index. But the issue of time lag is accentuated by the fact thatthe Case-Shiller Index is itself a moving average, so that the March 2010 indices (publishedin late May 2010) were themselves an average of information from January, February, andMarch 2010. Of course, the prices recorded in March 2010 and published in May wereagreed to at the beginning of their sale process, typically in January or February of that year,reflecting another unavoidable time lag.

This paper uses the Case Shiller Index as a general proxy for traditional views of the housingmarket. Other Home Price Indexes (HPIs) published by various public and private sources,exhibit similar characteristics and are lagging the actual current market by similar lengths oftime.

Real estate prices experience market inflection points from time to time. That is, homeprices may stop rising, reach a peak, and begin to fall or the other way round. The turns


3/44



may have short-term or long-term impact; that is, they may usher in a short or long period ofrising or falling prices. Clearly, lenders, investors, traders, and anyone interested in thehousing market will be affected by market turns and will desire to know as much about themas possible.

There is a clear need for real-time information about the level and direction of home prices

in a market, about current supply and demand characteristics, and peculiar market dynamics,particularly when market inflection points are considered.What is happening in the streetsin a housing market whether that market is hot or cold, whether buyers are eager orcautious, whether prices are rising, falling, or turning will be reflected and displayed inhousing price indices for that market, but only months after the fact. A housing market atits peak is characterized not only by leveling off in upward price change, but also by aslackening in buyer demand from eagerness to promptly pay almost any price to hesitation,caution, and even exhaustion; these shifts are noticeable by realtors and others on the scene

well before any index of sale prices is published.

This paper examines the power of using real-time, local dynamics as the new standard forunderstanding the housing market now and in the near future.

Altos Research Housing Market Statistics and Methodology

This paper references a subset of data available from Altos Research which tracks nearly 400statistics on the current state of local housing markets, describing what is happening in real-time:

Inventory (homes listed for sale on the market) Percent Price Decreased (percent of listings that have had their prices reduced) Median Asking Price of homes offered for sale Median Asking Price of New Listings (listings that have come on market in the last 7

days).

Median Price of Listings Absorbed (listings that have gone off market in the last 7 days) Mean Days on Market Median Days on Market Percent of Homes Relisted Year-over-Year % change in Asking Price Year-over-Year % Inventory Change Year-over-Year % change in Days on Market Year-over-Year % change in Percent Price Decreased

The present study will show that Altos indicators function as valuable real-time intelligenceof market conditions. Some Altos variables are positively correlated with price; others havenegative correlations; but in general, they lead closed transaction prices and home price

indices by two to eight months. This lead time provides a valuable advantage over waitingfor price indices to be published.

Even more important, Altos indicatorsand their rates of change or deltas will be seento serve as powerful leading indicators of market turns, being strongly correlated with therate of change of home prices themselves, with lead times of up to several months.

Therefore, knowing whether the Altos indicators are rising, falling, flat, or turning willprovide valuable information, months in advance about whether home prices will be rising,


4/44



falling, flat, or turning. This particularly valuable characteristic of the Altos variables will bean essential finding of this paper.

Further Information and Data

This paper was primarily authored by Christopher L. Cagan, Ph.D. for Altos Research LLC.For further information or to request time series data used in this white paper contact AltosResearch:

Scott SambucciVice president, Data [email protected] xt. 3

Michael SimonsenCo-founder [email protected] xt 4
mailto:[email protected]:[email protected]:[email protected]:[email protected]:[email protected]


5/44



Table of Contents

Executive Summary .............................................................................................................................. 1Watching active housing market statistics for clues about the future prices of homes. ............ 1Altos Research Housing Market Statistics and Methodology ........................................................ 2 Further Information and Data ............................................................................................................ 3

Table of Contents ................................................................................................................................. 4 Los Angeles, California: A Case Study .............................................................................................. 5Home Prices and Inventory ................................................................................................................ 6Home Prices and Percent Price Decreased ...................................................................................... 6Home Prices and Listing Prices .......................................................................................................... 7Home Prices and Days on Market ..................................................................................................... 9Changes in Home Prices and the Altos Leading Indicators ........................................................ 11Changes in Home Prices and Inventory ......................................................................................... 12Changes in Home Prices and Changes in Inventory ..................................................................... 17Changes in Home Prices and Percent of Listings with Price Decreased ................................... 20Changes in Home Prices and Changes in Percent of Listings with Price Decreased .............. 24Changes in Home Prices and Changes in Median Asking Prices ................................................ 28Changes in Home Prices and Changes in the Prices of New Listings ....................................... 32Changes in Home Prices and Percent Relisted .............................................................................. 36SUMMARY: Consensus, Currency, and Competitive Advantage .............................................. 40

Table of Figures .................................................................................................................................. 41 Data Series and Endnotes.................................................................................................................. 43


6/44



Los Angeles, California: A Case Study

The Los Angeles, California metropolitan area is one of the most important housing marketsin the country. It includes millions of residents ranging in economic status from very rich to

very poor. This study will examine the use of Altos variables and their rates of change inforecasting home prices, with particular emphasis on market turns. Because of limitations ontime and space, this paper will not look at other major metropolitan markets. However, thereis compelling evidence that the same methodology will work well in other majormetropolitan areas across the country, with the possibility of shorter or longer lead times indifferent markets. The Case-Shiller Home Price Index will be analyzed in relation to the

Altos indicators.

In investigating market turns, this study will look at three-month and six-month changes inhome prices as measured by the Case-Shiller Index, as correlated and compared with the

Altos indicators and particularly with three-month and six-month changes in thoseindicators. Since the rate-of-change or delta analysis is so important, we will also look atthe six-month changes in another index, the CoreLogic monthly Single Family Combined

index.

These three-month and six-month deltas in home price indices represent the rate of changeor velocity in home prices, expressing whether a market is rising, falling, or flat. In addition,the rates of change of the deltas themselves will hint at the convexity or concavity of homeprices, whether home prices are rising but slowing down in their ascent, or bottoming out asthey near the end of a bear phasepoints of inflection in a market. It will be shown that the

Altos variables and particularly their own changes or deltas serve as leading indicators months in advance of the rising and falling of home prices, and particularly of marketturns and points of inflection. As such, these variables make up a powerful and valuablesuite of on-the-ground leading indicators useful to anyone interested in the level anddirection of home prices.

The advantage of lead-time intelligence will be first observed by graphing the Altos variablesthemselves against the Case-Shiller Home Price Index. Later in the study the mathematicalanalysis will become more intense as we look at the relationship between rates of change ofthe different indicators with the application of forecasting market turns.


7/44



Home Prices and Inventory

Figure 1 below shows the active housing inventoryi graphed against the Case-Shiller indexii,both actual and reported. There is a two-month difference between the actual and reportedCase-Shiller indices; for instance, the March 2010 index was actually reported on May 25,2010.

Figure 1: Case-Shiller Index vs. Inventory

Inventory is negatively correlated with the strength or weakness of a market. A high

inventory is characteristic of a weak or weakening market, while a low inventory providesstrength and support to the market. In Figure 1, for example, we may compare peaks ininventory with bottoms in price.

There was a peak in inventory in the summer of 2008, which led by several months thecorresponding bottom in home prices in the spring of 2009. The low inventory in the

winter of 2009-2010 corresponds to a firmer period of home prices.

There is evidence for a correlative relationship here, even a leading relationship. However,the most powerful mathematical results will be seen later, in examining rates of change. Thepresent graph is simply an introduction.

Home Prices and Percent Price Decreased

The percentage of listings having price reductions is negatively related to the level of homeprices. A high level of price reductions is indicative of a weak or falling market, while a lowlevel of price reductions suggests a strong or rising market. Figure 2 shows the unique Altosindicator Percent Price Reductionsiii in comparison to the Case-Shiller index.


8/44



Figure 2: Case-Shiller Index vs. Percent of Homes with Price Reduction.Single Family Homes, 90-day rolling average

The inverse relationship is clear. A peak in Percent Price Decreased in the summer of 2009is associated with a price bottom at the same time or shortly afterward. A bottom in PercentPrice Decreased in early 2010 seems to be associated with a firming and a top in prices in thespring of that year.

As in the case of inventory, the strongest mathematical results will become apparent in theanalysis of rates of change.

Home Prices and Listing Prices

Figure 3 displays the relationship between the Case-Shiller index and median listing prices,median price of new listings, and median price of listings absorbed. The listing-price

variables are positively correlated with actual sale prices, as would be expected.


9/44


10/44



Home Prices and Days on Market

Figure 4 shows mean and median days on market vs. the Case-Shiller price index. In generalthe relationship is inverse. If homes take a long time to sell, their prices are likely to bereduced, and such conditions are characteristic of a weak market. On the other hand, in astrong market, homes sell quicklysometimes very quicklyafter they are listed.

Figure 4: Case-Shiller Index and Days on Market

The mean days on market (mean DOM) is a better indicator than median days on market.Mathematically, if a house sells in 100 days then we could model that 1/100 of that house issold per dayand this characteristic of rate-of-sale understanding lends itself to arithmeticand mathematics in a way that cannot be done by working with medians.

In general, the mean days on market has been increasing, corresponding to a more or lessweak housing market. The relationship is quite strong, as we see in the scatterplot of Figure5 below. Mean days on market is strongly correlated with the Case-Shiller Index, with notime lag used in plotting either variable (though we must remember that the Case-ShillerIndex itself is reported two months late).


11/44



Figure 5: Case-Shiller Index and Mean Days on Market

The correlation coefficient is a very powerful r = -0.875, with a level of significance of 3.4 E-13 (that is, 3.4 times 10 to the minus 13th power). The relationship is extremely strong.

Even more interesting is that the correlation using the mean days on market lagged threemonths only weakens to r = -0.849 with significance at 6.0 E-11 (still incredibly strong!).

This shift has gained three months of lead time at a very small cost in significance. It is morethan fair to say that mean days on market is a useful three-month leading indicator of saleprices. Remembering that the Case-Shiller index is reported two months late, we could saythat mean days on market is a useful five-month leading indicator in terms of actuallypredicting changes in sales price.

In future analysis we will see similar results: in many cases it will be possible to lengthen leadtimes at an acceptable cost in terms of correlation and significance, again illustrating theusefulness of the Altos variables in forecasting home price trends.


12/44



Changes in Home Prices and the Altos

Leading Indicators

The most powerful results of this paper, which provide the strongest evidence and supportfor the Altos variables as leading indicators for market trendsand particularly market turnswill appear when we look at the rates of change or deltas of home prices and of theAltos indicators.

To approximate the rate of change or velocity of home prices, we will use 3-month and 6-month changes in the Case-Shiller index, and also look at the 6-month changes in theCoreLogic index. These shorter-term deltas are better measures of the rates of price changethan a year-over-year, 12-month change.

Correspondingly, we will look at 3-month and 6-month deltas in the Altos variables. While12-month deltas are also available and easy to compute, these are less responsive inmeasuring the current rate of change of their underlying variables.

All deltas are computed by simple arithmetic and are not annualized. For instance, if avariable changes over three months from 100 to 105, this is reported as a delta of 0.05 or5%, rather than being annualized into a yearly rate of 0.20 or 20%. Deltas are reported in a

trailing fashion, so that a change of 5% from January to April is associated with the monthof April.

The following pages will provide many examples of this comparison of deltas, showing theeffectiveness of the Altos indicators in monitoring and leading home price changes, withparticular emphasis on market turns.


13/44



Changes in Home Prices and Inventory

A decreasing inventory of homes offered for sale is associated with a tightening housingmarket, as these homes sell quickly and for high prices, while an increasinginventory isassociated with a weakening market. This relationship is particularly evident when inventoryis studied in connection with the rate of change or deltas in home prices.

Figure 6 shows the course of inventory graphed together with the six-month change (notannualized) in the Case-Shiller index. To recognize the two-months lag in the reporting ofthe index, Figure 6 displays both the actual value of the delta of the Case-Shiller index andthen, lagged by two months, the actual reporting time of the information.

Figure 6: Six month change in Case-Shiller Index and Available Inventory

The relationship between inventory and the rate of price changes is clearly inverse andclearly strong. A peak in inventory in November 2007 foreshadowed a bottom in the

velocity of home prices in April 2008, reported in June 2008. It must be remembered that abottom in this velocity (lowest negative value) does not mean a bottom in prices themselvesbut rather a time of fastest descentin prices (prices falling the fastest), which itself is usefulinformation. This represents a point of inflectionin home prices; before April 2008 the bearmarket was accelerating, while after April 2008 the decline was slowing down; hence we haveinsight into the second derivative (acceleration or deceleration) in home prices.

Similarly, a second peak in inventory in July 2008 came in advance of a correspondingsecond bottom in the velocity of prices in February 2009.

Inventory declined from February 2009 until January 2010, roughly corresponding to anincrease in the rate of change of home prices. Home prices had been declining at a 6-monthrate of 10% or so in the spring and summer of 2009, stopped declining by August 2009 (therate of change crossed the zero line and moved from negative to positive between July and


14/44



August 2009), and then entered positive territory (home prices rising, a rate of change abovezero).

Of course, a zero level of home price change corresponds to the flat characteristic of a peakor bottom in home prices in other words, to a market turn from bear to bull or from bullto bear. Thus we see that inventory is a useful leading indicator in relation to market turns.

Looking more deeply, a bottom in the rate of change of prices corresponds to an inflectionpoint from accelerating decline towards decelerating decline and market firming, while apeak in the rate of change of prices corresponds to an inflection point from acceleratingprice growth towards a time of slowinganother type of market shift. Again we see thatinventory is a useful leading indicator in suggesting future changes in the tenor and characterof a housing market. Those using inventory as a leading indicator in real time will knowmonths in advance how a market is likely to perform, rather than having to wait six (oreight!) months for the actual numbers to be reported.

The correlation between the rate of change in home prices and inventory is statistically verypowerful, underlining the usefulness of this variable as a leading indicator. Figure 7 is a

scatter plot of the six-month change in the Case-Shiller index against inventory lagged by 5months.

Figure 7: Case-Shiller Index, six month change and Inventory, lagged five months

The relationship, clearly inversely correlated, is statistically very powerful:

Correlation r = -0.914Significance = 1.0 E-13 (1.0 times 10 to the -13th power)

This high degree of correlation is operative five months in advance, since the inventory timeseries was lagged by five months. Inventory, therefore, functions as an extremely strongindicator of change in home prices, with a lead time of five months (seven months in


15/44



relation to the reporting of the price index and thus to the availability of knowledge about itsdeltas).

Increasing the lead time to seven months (that is, lagging the inventory data series by sevenmonths) weakens the correlation only slightly:

Correlation r = -0.873Significance = 3.3 E-11 (3.3 times 10 to the -11th power)

Here the analysis gains an additional two months of lead time, while weakening thecorrelative relationship only slightly, as the correlation still possesses an overwhelmingstatistical significance.

The relationship between inventory and price velocity is also apparent in looking closer tocurrent time, studying the three-monthchange in the Case-Shiller index rather than the six-monthchange. Please refer to Figure 8 below.

Figure 8: Three-month change in Case-Shiller Index and Inventory

Although the three-month rate of change is more volatile (and more current) than its six-month counterpart, the same features of tops and bottoms, leads and lags are seen in Figure8 as previously appeared in Figure 6. Figure 9 displays the scatter plot of the three-monthchange in prices with respect to inventory, lagged four months (the lag is shorter with thetime span in home price changes being shorter).


16/44



Figure 9: Case-Shiller Index, three month change and Inventory, lagged 4 months

Again the relationship is inverse and statistically very significant. The results are:

Correlation r = -0.882 (only minor damage due to the volatility of 3 month changes)Significance = 2.5 E-12.

In this case inventory functions as a four-month leading indicator relative to the Case-Shillerprice index and as a six-month leading indicator relative to the actual reporting of that index.

Adding an extra two months of lead time (using inventory lagged by six months) is gained atonly a very small cost in statistical significance:

Correlation r = -0.755Significance = 6.8 E-7.

The usefulness of inventory is not limited to the Case-Shiller index. Figure 10 shows thelevel of inventory in relation to the six-month change in the CoreLogic home price index.


17/44



Figure 10: Six-month Change in CoreLogic Index and Inventory

Just as with the Case-Shiller index, peaks in inventory generally come before bottoms in thesix-month rate of change in the CoreLogic index. Figure 11 is the corresponding scatter plot.

Figure 11: CoreLogic Index, six-month change and Inventory, lagged six months

Using inventory as a six-month leading indicator (lagged 6 months) yields r = -0.864 and asignificance of 9 E-11. Adding an additional two months of lead time for a total of eightmonths produces a correlation of r = -0.726 and a significance of 3.8 E-6 (that is,0.0000038).


18/44



Changes in Home Prices and Changes in Inventory

While inventory itself serves as a useful leading indicator of the rate of change of homeprices, the same is also true of the velocity or rate of change in inventory. This rate ofchange reflects both how fast new listings are entering the market and how fast homes arebeing removed from the market (new listings increase the inventory while sold homes reduceit). In this section we will study the rate of change in home prices in relation to the rate ofchange in inventory as measured over a corresponding time span.

Figure 12 displays the six-month change in the Case-Shiller index and the six-month changein inventory. Since the change or delta in any variable is more volatile than the variable itself,it is not surprising that the change in inventory is somewhat jagged. However, a similar lead-lag relationship is generally observed.

Figure 12: Six-month change in Case-Shiller Index and six-month change in inventory

The relationship is better illustrated in the scatter plot of Figure 13 below.


19/44



Figure 13: CSI six-month change and Inventory 6 month change, lagged seven months

The relationship is not perfectly linear but is definitely apparent. The six-month change ininventory leads the six-month change in the price index by seven months (by nine months

when compared to the index being actually reported). The statistical results are:

Correlation r = -0.760Significance 6.8 E-6 (that is, 0.0000068).

The delta-to-delta correlation is particularly useful in forecastingmarket turns. If the rate ofchange of inventory crosses the zero line (moving from positive to negative or the other wayround), this could serve as a seven-month leading indicator of a change in the direction ofhome prices.

Some of the correlation is related to the common effect of seasonality. In the Los Angelesarea the most popular buying time is in the late winter, spring, and early summer. Thehousing market is slower in the fall and early winter. But the correlative relationship is stillclear and useful in its own right, even recognizing that some of the motion in both variablesactually reflects the underlying common factor of seasonality.

The lead and lag times in this study are not to be considered as absolute. In a slow-movingmarket where homes may take a year or more to sell, longer lead and lag times are to beexpected. On the other hand, in a fast-moving market with low inventory and propertiesselling rapidly, the situation would be much more sensitive, and shorter lead times wouldlikely be observed.

The same relationships, though not so precise, appear in studying the six-month changes inthe CoreLogic index rather than the Case-Shiller index. Please refer to Figures 14 and 15below.


20/44



Figure 14: Six-month change in CoreLogic HPI and six-month change in Inventory

Figure 15: Six-month change in CoreLogic HPI and six-month change in Inventory, lagged

seven months

Again the relationship is not quite linear but is still useful, with r = -0.648 and significance of3.4 E-4 (that is, 0.00034) using a seven-month lead/lag time.


21/44



Changes in Home Prices and Percent of Listings with Price Decreased

There is an inverse relationship between the rate of change in home prices and the percentof listings with price reductions. A high level of price reductions suggests a weak market,

while a low level of reductions is associated with a strong market. Figure 16 graphs the six-month change in the Case-Shiller index and the percent of listings with price decreased.

Figure 16: Six-month change CSI and Percent of Homes with Price Reduction

The peak in Percent Price Decreased in January 2008 led the April 2008 bottom in the rateof change of the Case-Shiller index by three months. The February-March 2009 peak inPercent Price Decreased and its subsequent dip did not lead the corresponding bottom inthe velocity of the price index but were instead concurrent with it.


22/44



The scatter plot of Figure 17 below uses a two-month lag in Percent Price Decreased, givingPercent Price Decreased a two-month lead over the six-month rate of change of the priceindex (and a four-month lead over when the information is available).

Figure 17:Case-Shiller Index, 6 months change, and Percent Price Decreased, lagged 2months

The relationship, although not exact, is statistically significant, with a correlation of r = -0.653 and significance = 3.8 E-5 (that is, 0.000038).

An improvement in significance is gained at the sacrifice of lead time. If Percent PriceDecreased is not lagged at all, the correlation strengthens to r = -0.835 and significance = 1.5E-9.

Percent Price Decreased is not absolute and compelling as a leading indicator, but it certainlyhelps to suggest the direction in which prices are likely to move. Percent Price Decreased, inconjunction with the other Altos variables, is part of a powerful consensus pointing the wayto the direction and turning of home prices.


23/44



Figure 18 below looks at the shoter-term, more nearly current, three-month change in theCase-Shiller index together with Percent Price Decreased.

Figure 18:Case-Shiller Index, 3 months change, and Percent Price Decreased

The scatter plot of Figure 19 is built with no lag at all in Percent Price Decreased.

Figure 19: Case-Shiller Index, 3 months change, and Percent of Homes with Price Reduction

The relationship, while not perfectly linear, is certainly useful. The correlation is r = -0.682with significance = 4.6 E-6. Adding two months of lead time weakens the correlation to r =-0.465 with significance = 0.004 (still highly significant in any statistical analysis); and this

variable is actually four months ahead of the reportage of the Case-Shiller price index andthus also of the knowledge about its rate of change.


24/44



Figures 20 and 21 display the corresponding results for the six-month change in theCoreLogic price index.

Figure 20: CoreLogic HPI, 6 months change, and Percent of Homes with Price Reductions

Figure 21: CoreLogic HPI, 6 months change, and Percent of Homes with Price Reductions,lagged 2 months

Again the relationship is useful although not absolutely linear. The correlation is r = -0.631and the significance is 2.8 E-5. Giving up the two months of lead time (thus using no lag)strenghens the correlation to r = -0.768 and the significance to 1.2 E-8.


25/44



Changes in Home Prices and Changes in Percent of Listings with Price

Decreased

The inverse relationship between home prices and reductions in listing prices is most clearly

seen by looking at the deltas of both of these variables. Figure 22 graphs the six-monthchange in the Case-Shiller index with the six-month changes in Percent Price Decreased.

Figure 22: Case-Shiller Index, 6 months change, and Percent of Homes with PriceReductions, 6 months change

Figure 23 is the corresponding scatter plot, with the six-month changes in Percent PriceDecreased being lagged by four months.

Figure 23: Case-Shiller Index, 6 months change, and Percent of Homes with PriceReductions, 6 months change, lagged 4 months


26/44



Although the relationship is not absolutely linear it is certainly strong, with correlation r = -0.850 and significance = 5.3 E-9. The change in Percent Price Decreased has a four-monthslead on the change in the price index (six months lead over the actual knowledge of thatchange).

This correlative relationship between two rates of change is particularly useful in forecasting

market turns before they actually happen. A shift in the velocity of Percent Price Decreasedfrom positive to negative may signal an (inverse) change in the direction of prices fromnegative to positive, with a lead time of four (functionally six) months. Interpreting this interms of prices instead of rates of change, if fewer listings have their prices reduced, this islikely to suggest a market of rising prices. Knowingor even seriously surmising that amarket turn or point of inflection is in the offing is extremely valuable to anyone interestedin the housing market.

Looking from a shorter-term, more nearly current perspective, Figure 24 graphs the three-month change in the Case-Shiller index and the three-month change in Percent PriceDecreased.

Figure 24:Case-Shiller Index, 3 months change, and Percent of Homes with PriceReductions, 3 months change

The relationship is again inverse, though more volatile than in the case of a six-month timelapse.


27/44



Figure 25 below shows the corresponding scatter plot, again with a four-month lag appliedto the three-month change in Percent Price Decreased.

Figure 25: Case-Shiller Index, 3 months change, and Percent of Homes with PriceReductions, 3 months change, lagged 4 months

Again the relationship is strong enough to provide useful market intelligence with a four-months time lead (six months on the reported index), particularly if it appears in consensus

with other Altos indicators. The correlation is r = -0.784 and the significance is 1.1 E-7. Ifthe rate of change in Percent Price Decreased moves from positive to negative (fewerproperties having their prices reduced), a firming or strengthening of market prices is likelyto come within a few months; and on the other hand, if the velocity of Percent PriceDecreased moves from negative to positive (more properties having their prices reduced),this is likely to suggest a subsequent weakening of the housing market.


28/44



A similar relationship is seen in reference to the six-month change in the CoreLogic index.Please refer to Figures 26 and 27 below.

Figure 26: CoreLogic index, 6 months change, and Percent of Homes wth Price Reductions,6 months change

Figure 27: CoreLogic index, 6 months change, and Percent of Homes with Price Reductions,6 months change, 5 months lag

Here the correlation is r = -0.718 and the significance is 1.7 E-5 (that is, 0.000017).


29/44



Changes in Home Prices and Changes in Median Asking Prices

There is an obvious relationship between asking prices and actual sale prices. The same istrue of their deltas (changes). Figure 28 graphs the six-month change in the Case-Shillerindex with the six-month change in median asking price. Figure 29 is the correspondingscatter plot with a two-month lag applied to the delta of the median asking price.

Figure 28: Case-Shiller Index, 6 months change, and Median Asking Price, 6 months change

Figure 29: Case-Shiller Index, 6 months change, and Median Asking Price, 6 months change,2 months lag


30/44



The correlation is r = 0.799 and the significance is 6.9 E-8. The lead time of two months isactually four months relative to release of the Case-Shiller index. The relationship betweenthe two variables is not absolutely linear, but it is sufficiently strong to make the medianasking price and its rate of change something well worth knowing about. If the change inmedian asking price crosses the zero line, shifting from negative to positive (asking pricesstart rising instead of falling), this is a likely signal of a corresponding positive market turn in

actual sale prices.

Figure 30 gives the corresponding displays for the three-month change in the Case-Shillerindex and the three-month change in median asking price.

Figure 30: Case-Shiller Index, 3 months change, and Median Asking Price, 3 months change

Note that in the summer and fall of 2009 the velocity of the median asking price peaked inadvance of the peak and subsequent drop in the rate of change of the Case-Shiller indexitself. That is, the heat or fury of rises in asking prices (the rate of increase in askingprices) peaked in intensity (fastest rate of increase) shortly before the same thing happened

with respect to actual sale prices.

The rate of change of median asking price crossed the zero line and changed from positiveto negative in December 2009, as asking prices peaked and began to fall. This led by threemonths the transition from rising to falling in the Case-Shiller index (three-month rate ofchange moved from positive to negative) in March 2010, although this was not actuallyknown until late in May of that year. It was indeed worth knowing what was happening toasking prices!


31/44



The scatter plot of Figure 31 uses a two-month time lag applied to the median asking price.

Figure 31: Case-Shiller Index, 3 months change, and Median Asking Price, 3 months change,

2 months lag

The relationship for the more volatile shorter-term deltas is weaker, with r = 0.642 andsignificance of only 4.3 E-5 (or 0.000043). As before, the relationship is not absolute but iscertainly evident enough to make the change in median asking price worth knowing incurrent time.


32/44



Figure 32 displays the results for the six-month change in the CoreLogic index.

The rate of change of the median asking price reached a top in the fall of 2009, slightlyleading the change in the CoreLogic price index. That is, the rate of increase in medianasking price reached its maximum (hottest rate of increase) in October 2009, one monthbefore the fastest rate of increase in the CoreLogic index, in November 2009. The six-month

change in median asking price crossed the zero line and shifted from positive to negative(declining asking prices) in January 2010, again one month ahead of the corresponding shiftin the rate of change of the CoreLogic index in February.

The six-month change in median asking price reached a bottom in January 2009, twomonths before the corresponding bottom in the rate of change of the CoreLogic index inMarch of that year.

Figure 32: CoreLogic Index, 6 months change, and Median Asking Price, 6 months change


33/44



Figure 33 shows the scatter plot. A two-month lag has been applied to the change in medianasking price.

Figure 33: CoreLogic Index, 6 months change, and Median Asking Price, 6 months change, 2months lag

The correlation of r = 0.821 is quite strong, with significance = 1.5 E-8. Again, the changesin median asking price were well worth knowing about, giving a two-months signal on what

was likely to happen to the changes in home prices themselves.

Changes in Home Prices and Changes in the Prices of New Listings

The asking prices of new listings particularly reflect the current, real-time perceptions ofsellers about how prices have recently changed and are expected to change. There is apositive relationship between changes in the median price of new listings and the changes inactual sale prices, as seen in Figure 34 below.


34/44



Figure 34: Case-Shiller Index, 6 months change, and Median Price of New Listings, 6 monthschange

The bottom in rate of change of the median price of new listings (time of fastest decrease)occurred in January 2009, one month in advance of the time of fastest decrease in the Case-Shiller index and three months before that information was actually available. In July 2009the rate of change of median price of new listings crossed the line from negative to positive(median price of new listings stopped falling and began to rise), again one month before thecorresponding event with respect to actual sale prices. The time of fastest increase in price ofnew listings was the strong peak of October 2009, again one month before the time offastest sale price increases (hottest price rises) in November of that year.

Figure 35 shows the corresponding scatter plot. A lag of two months has been applied to therate of change of median asking price.

Figure 35: Case-Shiller Index, 6 months change, and Median Price of New Listings, 6 monthschange, 2 months lag

Although there are a few points off the trend line, the overall positive relationship is clear.The correlation is a very high r = 0.905 and the significance is 1.3 E-09, almost down to onepart in a billion. What happens in the velocity of the price of new listings is very likely to


35/44



signal what will happen to the rate of changethe direction of motionin actual sale prices.Here there is a two-month lag, but this is actually four months due to the delay in reporting.

When the time interval is shortened to a more current but less stable three months, therelationship is less obvious. Please refer to Figures 36 and 37.

Figure 36: Case-Shiller Index, 3 months change, and Median Price of New Listings, 3 monthschange

Figure 37: Case-Shiller Index, 3 months change, and Median Price of New Listings, 3 monthschange, 3 months lag

Even here, the relationship is still statistically strong, with r = 0.485 and a significance levelof 0.012.


36/44



With a time interval of six months, the rate-of-change relationship is even stronger when theCoreLogic index is studied instead of the Case-Shiller index. Please see Figure 38 below.

Figure 38: CoreLogic Index, 6 months change, and Median Price of New Listings, 6 monthschange

The six-month velocity or change in median price of new listings clearly led the six-monthrate of change in the CoreLogic index at least three times corresponding to two points ofinflection and one market bottom in home prices themselves.

1. In reaching a low in January 2009 (fastest rate of decrease in new listing prices)2. In crossing the zero line from negative to positive (moving from falling new listing

prices to rising new listing prices) in July 2009a market bottom.3. In reaching a top in October 2009 (fastest rate of increase in new listing prices).


37/44



Figure 39 displays the corresponding scatter plot.

Figure 39: CoreLogic Index, 6 months change, and Median Price of New Listings, 6 monthschange, 1 month lag

The relationship has only a one-month lead time, but is extremely powerful with acorrelation of r = 0.940. The significance level is 3.1 E-12, or about three parts in a trillion.

It is clear that what happens to the velocity of the median price of new listings will soon bereflected in the rate of change of sale prices themselves.

Changes in Home Prices and Percent Relisted

If many properties are relisted (typically at lower prices), this is indicative of a weak ordeclining market, while fewer properties are relisted in a strong or rising market. Therelationship between the two is thus inverse.

Figure 40: Case-Shiller Index, 6 months change, and Percent Relisted


38/44



Figure 40 displays the inverse nature of the relationship. It will be helpful to look at thescatter plot of Figure 41.

Figure 41: Case-Shiller Index, 6 months change, and Percent Relisted, lag 2 months

The relationship between Percent Relisted and the six-month delta in the Case-Shiller priceindex (with a two months lead over the actual index and a four months lead over thereported index) may not look powerful but it is statistically real, with r = -0.707 andsignificance = 4.2 E-6. We can get an extra two months of lead time by lagging PercentRelisted by four months, at the bearable cost of reducing r to -0.632 and significance toonly 8.0 E-5 (or 0.00008).

When the time span is shortened to three months the relationship is more volatile, as seen in

Figures 42 and 43.

Figure 42: Case-Shiller Index, 3 months change, and Percent Relisted


39/44



Figure 43:Case-Shiller Index, 3 months change, and Percent Relisted, lag 2 months

In this case the correlation is still a powerful r = -0.563 and the significance is 3.5 E-4 (or

0.00035), still a strong relationship.

Figure 44 looks at the six-month change in the CoreLogic index in relation to PercentRelisted.

Figure 44: CoreLogic Index, 6 months change, and Percent Relisted

Again the relationship is clearly inverse.


40/44



Figure 45 is the corresponding scatter plot, with a two month lag applied to Percent Relisted.

Figure 45: CoreLogic Index, 6 months change, and Percent Relisted, 2 months lag

The correlation is a strong -0.800 and the level of significance is 2.9 E-9, around three partsin a billion.


41/44



SUMMARY: Consensus, Currency, and

Competitive Advantage

We have seen that the Altos variables and their deltas have powerful leading relationships with home prices and with the changes in those prices, with particular reference toforecasting market turns such as tops, bottoms, and points of inflection. Each relationshipexhibited in this paper is statistically significant. While no correlation is absolute, the total setof Altos tools, taken together, provide a powerful community of consensus about what ishappening and what is likely to happen.

The present study illustrates the importance ofcurrency in knowledge. While home priceindices by their very nature are only known well after the actual sales prices are agreed upon,the Altos variables provide real-time information, in advance of sales, about current marketconditions. Indicators related to inventory, new listings, and asking prices stand logicallyand chronologically in advance of the closing of escrows, in advance of the recording of

sale prices, and even farther in advance of the collection and reporting of home price data inthe form of an index. This advantage in time is particularly important in relation to peaks ortroughs in prices, and in relation to points of inflection when the pace of price rises ordeclines picks up or slows down.

Currency of information real-time knowledge can be valuable to lenders, investors,traders, builders, realtors, and everyone involved in the housing market. It would have been

very helpful in 2007 to have leading information about the price peak of that year, and itwould have been very helpful in 2008 to know in advance even if only in the form of asuggestion something about the power of that years housing crash. In any year, it ishelpful to know about price dips, peaks, rises and declines.

The value of current and leading information is all the greater in the context of competitiveadvantage. Investors, traders, and all market participants will gain an advantage in havingsuperior market intelligence particularly current and leading intelligence that theircompetitors lack. On the other hand, investors, traders and market participants that basetheir portfolio decisions on lagging market informationwill be significantly disadvantagedin an environment where their competitors have real-time knowledge.

Knowledge is power. Knowledge is money. The Altos indicators are an aid to knowledge.


42/44



Table of Figures

Figure 1: Case-Shiller Index vs. Inventory ........................................................................................ 6Figure 2: Case-Shiller Index vs. Percent of Homes with Price Reduction. ................................. 7Figure 3: Case-Shiller Index and Asking Prices. Vertical bars indicate inflection points ofactive market new listings and, later, for the reported data of the Case-Shiller Index .............. 8Figure 4: Case-Shiller Index and Days on Market ........................................................................... 9Figure 5: Case-Shiller Index and Mean Days on Market .............................................................. 10Figure 6: Six month change in Case-Shiller Index and Available Inventory ............................. 12Figure 7: Case-Shiller Index, six month change and Inventory, lagged five months ............... 13Figure 8: Three-month change in Case-Shiller Index and Inventory ......................................... 14Figure 9: Case-Shiller Index, three month change and Inventory, lagged 4 months ............... 15Figure 10: Six-month Change in CoreLogic Index and Inventory ............................................. 16Figure 11: CoreLogic Index, six-month change and Inventory, lagged six months ................ 16Figure 12: Six-month change in Case-Shiller Index and six-month change in inventory ....... 17

Figure 13: CSI six-month change and Inventory 6 month change, lagged seven months ...... 18Figure 14: Six-month change in CoreLogic HPI and six-month change in Inventory............ 19Figure 15: Six-month change in CoreLogic HPI and six-month change in Inventory, laggedseven months ....................................................................................................................................... 19Figure 16: Six-month change CSI and Percent of Homes with Price Reduction ..................... 20Figure 17:Case-Shiller Index, 6 months change, and Percent Price Decreased, lagged 2months .................................................................................................................................................. 21Figure 18:Case-Shiller Index, 3 months change, and Percent Price Decreased ........................ 22Figure 19: Case-Shiller Index, 3 months change, and Percent of Homes with Price Reduction................................................................................................................................................................ 22Figure 20: CoreLogic HPI, 6 months change, and Percent of Homes with Price Reductions................................................................................................................................................................ 23

Figure 21: CoreLogic HPI, 6 months change, and Percent of Homes with Price Reductions,lagged 2 months .................................................................................................................................. 23Figure 22: Case-Shiller Index, 6 months change, and Percent of Homes with PriceReductions, 6 months change ........................................................................................................... 24Figure 23: Case-Shiller Index, 6 months change, and Percent of Homes with PriceReductions, 6 months change, lagged 4 months ............................................................................ 24Figure 24:Case-Shiller Index, 3 months change, and Percent of Homes with PriceReductions, 3 months change ........................................................................................................... 25Figure 25: Case-Shiller Index, 3 months change, and Percent of Homes with PriceReductions, 3 months change, lagged 4 months ............................................................................ 26Figure 26: CoreLogic index, 6 months change, and Percent of Homes wth Price Reductions,6 months change ................................................................................................................................. 27Figure 27: CoreLogic index, 6 months change, and Percent of Homes with Price Reductions,6 months change, 5 months lag ........................................................................................................ 27Figure 28: Case-Shiller Index, 6 months change, and Median Asking Price, 6 months change................................................................................................................................................................ 28Figure 29: Case-Shiller Index, 6 months change, and Median Asking Price, 6 months change,2 months lag ......................................................................................................................................... 28Figure 30: Case-Shiller Index, 3 months change, and Median Asking Price, 3 months change................................................................................................................................................................ 29


43/44



Figure 31: Case-Shiller Index, 3 months change, and Median Asking Price, 3 months change,2 months lag ......................................................................................................................................... 30Figure 32: CoreLogic Index, 6 months change, and Median Asking Price, 6 months change................................................................................................................................................................ 31Figure 33: CoreLogic Index, 6 months change, and Median Asking Price, 6 months change,2 months lag ......................................................................................................................................... 32

Figure 34: Case-Shiller Index, 6 months change, and Median Price of New Listings, 6months change .................................................................................................................................... 33Figure 35: Case-Shiller Index, 6 months change, and Median Price of New Listings, 6months change, 2 months lag ........................................................................................................... 33Figure 36: Case-Shiller Index, 3 months change, and Median Price of New Listings, 3months change .................................................................................................................................... 34Figure 37: Case-Shiller Index, 3 months change, and Median Price of New Listings, 3months change, 3 months lag ........................................................................................................... 34Figure 38: CoreLogic Index, 6 months change, and Median Price of New Listings, 6 monthschange ................................................................................................................................................... 35Figure 39: CoreLogic Index, 6 months change, and Median Price of New Listings, 6 monthschange, 1 month lag ............................................................................................................................ 36

Figure 40: Case-Shiller Index, 6 months change, and Percent Relisted ..................................... 36Figure 41: Case-Shiller Index, 6 months change, and Percent Relisted, lag 2 months ............ 37Figure 42: Case-Shiller Index, 3 months change, and Percent Relisted ..................................... 37Figure 43:Case-Shiller Index, 3 months change, and Percent Relisted, lag 2 months ............. 38Figure 44: CoreLogic Index, 6 months change, and Percent Relisted ....................................... 38Figure 45: CoreLogic Index, 6 months change, and Percent Relisted, 2 months lag .............. 39


44/44


Data Series and Endnotes

i Inventory: total homes listed for sale in Los Angeles Metropolitan Statistical Area, 90-day rollingaverage, single family homes. Altos Research time-series data product number: ZR.ZRINV.90.SF.A-Los

Angeles-Long Beach-Santa Anaii Case-Shiller Index: Los Angeles S&P Case Shiller Index CSX.LAiii Percent Price Reductions: Percent of homes currently on the market that have had price reductions.Time-series data product number: ZR.PPD.90.SF.A- Los Angeles-Long Beach-Santa Ana

Documents

Altos Research: Leading Indicators of Home Prices (Sept 2010)